I'm not positive, but it's my impression that push, pop, shift, and unshift are all implemented as either thin wrappers around splice or thin wrappers around whatever underlies splice. So I think that the most possible difference is a single stack-frame (call unshift, and then unshift calls splice, as opposed to calling splice directly)... but I'm not even sure that that isn't optimized away (as, like, a macro or some such in the underlying C code). To know for sure, I'd have to dig through the C source for the perl core, which I don't really feel like doing right now. ;-)